On Sun, Jan 12, 2020 at 11:16 AM Kerim Aydin via agora-official <agora-offic...@agoranomic.org> wrote: > =============================== CFJ 3787 =============================== > > Murphy initiated an election for Prime Minister on or about Sun, > Dec 22 2019 (18:48:16) UTC. > > ========================================================================== > > Arbitor's Evidence: > > This is concerning whether the following message was a public message: > > https://mailman.agoranomic.org/cgi-bin/mailman/private/agora- > official/2019-December/013298.html > > ==========================================================================
Based on my timeline of events [1], the message in question was sent a week after Gmail had started rejecting most or all list messages, and a few hours before I started changing settings in an attempt to address the situation. Convenient, since my tweaks muddled things for the next few days even more than they had been. One potential complication is that, due to Murphy's email domain's strict DMARC policy, e is the sole player who already had the From header of eir messages rewritten to agoranomic.org – the same measure that I eventually enabled for all senders, in order to end the Troubles for now. [2] However, when I did that, I also added SPF and DKIM records for agoranomic.org and enabled outgoing message signing, since Gmail's error message had complained about missing "authentication information". That appears to have made the difference. In the case of Murphy's earlier message, it doesn't seem that From rewriting alone was enough to satisfy Gmail. The message doesn't appear in my own Gmail account, and if I'm reading my server logs correctly (they're unfortunately not clear about which message is which, something I've since improved), Gmail did reject all copies of the message with the same error as everyone else's messages. So, the facts are: - The mailing list server received Murphy's message without issue, shortly after e sent it. - At that time it was added to the archive, and queued for delivery to each non-digest subscriber. - It appears that delivery to everyone *not* using Gmail succeeded as usual. - But the majority of players use Gmail, whose mail server rejected most or all copies of the message. In other words, the message successfully left Murphy's "technical domain of control" (a term from past judgements), but never entered most players' technical domains of control. This is fairly close to the situation in CFJ 1905 [3]: > comex cites CFJ 1314, which appears to set the precedent that a > message sent to a Public Forum is public, even if it is sent from > an unsubscribed address and thus fails to be relayed. However, > Rule 478 did and still does require public messages to be sent > /via/ a Public Forum, or to all players with a clear designation > of intent to be public. (In other words, the precedent implied by > CFJ 1314 was incorrect, but was not recognized as such until now.) > > Based on common sense and the best interests of the game, I > interpret "sent via a Public Forum" as requiring the Public Forum > in question to re-send the message to a reasonably large subset of > the set of all persons who have reasonably arranged to receive > messages via that Public Forum. Thus, a message that is initially > blocked, but subsequently approved by the Distributor and re-sent, > would qualify as public. (The "reasonably large subset" clause is > to avoid disqualifying messages due to a hiccup involving just one > or a few subscribers.) The one difference is that in that case, the mailing list made no attempt to relay the message to others (because it was sent from a non-subscribed address), whereas in the present case the list tried but was rebuffed. Now... the judgement makes a distinction between messages "sent to" a forum and "sent via" the forum; by the same principle, you could argue that the copies of the message were "sent to" Gmail but not "sent via" Gmail. And yet, by the literal wording of the judgement, the Public Forum is only required to re-send the message "to" a reasonably large subset of subscribers! But that's being overly literal. Or even if you do interpret the judgement that way, that just means the judgement is wrong in this specific case, as it failed to consider the possibility of re-sent messages being rejected. The same R217 factors cited in the judgement, "common sense and the best interests of the game", clearly counsel in favor of requiring the re-sending to succeed. On the other hand: what if, as a scam, a significant fraction of players arrange to intentionally block messages from the forum? Could they thereby prevent all list messages from being public? I'd say no, as that would mean they had no longer "reasonably arranged to receive messages", no different from if they had unsubscribed. On the other hand, merely using an email provider with an overzealous spam filter (Gmail) is not tantamount to unsubscribing. After all, the delivery issues *were* fixable (even if the From line is less useful for the time being). And it clearly wouldn't be in the interests of the game to disregard deliverability for the majority of players! To sum up, Murphy's message was not a public message because it was not "sent via" the list to a large-enough subset of subscribers. Therefore no election was initiated at that time. FALSE. [1] https://mailman.agoranomic.org/cgi-bin/mailman/private/agora-discussion/2019-December/056067.html [2] Incidentally, I still see that as a temporary measure. I planned to wait a few weeks before trying to turn it off again, and it's now been four weeks, so it's around time. But rather than just flipping the switch and risking another outage, I plan to hack up my mail server to *automatically* handle rejections by retrying with a rewritten From address. That way, if it turns out that Gmail hasn't resumed trusting my server enough to 'forge' messages from other domains, or if it decides to distrust it again in the future, messages will still get through with the rewritten address. It may be another week or two before I get around to implementing this. [3] https://faculty.washington.edu/kerim/nomic/cases/?1905